Sensitivity of Urinary Toxicant Co-Exposure Patterns to Demographic Adjustment and Marker Type: A Methodological Analysis.

Patients presenting with conditions attributed to environmental exposures face complex, multi-toxicant burdens, yet the stability of multivariate toxicant patterns under demographic adjustment remains poorly evaluated. This study assessed the sensitivity of principal component analysis (PCA) structures to demographic confounding and variable composition in 551 patients (aged <1-86 years) with environmentally related conditions. Nine urinary biomarkers were analysed using PCA with Varimax and Promax rotation on both raw data and residuals adjusted for age and sex. Among the identified patterns, the solvent/industrial cluster (xylene and styrene metabolites) was the most stable, persisting across both raw and residual analyses regardless of rotation method. However, overall component structures were sensitive to preprocessing: the clustering pattern of the herbicide 2,4-dichlorophenoxyacetic acid shifted markedly after demographic adjustment, illustrating indirect confounding whereby demographic effects on co-variables altered apparent biomarker associations. Notably, inclusion of an endogenous mitochondrial marker (tiglylglycine) alongside exogenous toxicant biomarkers produced Heywood cases (loadings > 1), violating factor analysis assumptions and indicating that mixing exposure and response variables destabilises the model. Cumulative variance explained was modest, consistent with the weak inter-biomarker correlations observed (KMO ≈ 0.52). These findings do not support the identification of robust, demographically stable clustering patterns in this cohort. Instead, they demonstrate that PCA-derived structures in heterogeneous clinical data are vulnerable to demographic confounding, variable selection and marker type, and caution against interpreting transient clustering patterns as definitive exposure signatures without rigorous validation. It should be noted that the absence of a non-clinical matched control group means that whether the identified co-exposure signatures are distinctive to symptomatic patients or reflective of general population exposure patterns cannot be determined; future studies should incorporate matched controls to address this directly.